Reproducibility data and code for "A Systematic Literature Review of AI-Based Decision Support in Higher Education (2015-2026)"

Description

This dataset contains the underlying research data for the systematic literature review "A Systematic Literature Review of AI-Based Decision Support in Higher Education (2015-2026)." It comprises the bibliographic corpus retrieved from the open OpenAlex database, the record-level title/abstract screening decisions, and the computed descriptive (bibliometric) results. A structured search (12 queries; 2015-2026; English; journal and conference articles) retrieved 1,190 unique records, of which 1,105 with retrievable abstracts were screened against predefined inclusion/exclusion criteria, yielding 531 included studies and a prioritized core of 95 studies analysed in depth. Files are organised in three folders: - research/: the search log, the retrieved corpus (titles, years, abstracts, authors, venues, author countries, citation counts), and bibliographic metadata for the 95 core and 30 background references. - screening/: the complete record-level screening decision log (include/exclude decision, decision-support category, principal AI method, relevance rating, and reason) and the included (n=531) and core (n=95) study sets. - bibliometrics/: all computed descriptive results (publication trends, decision-support categories, AI method distribution, author geography, venues, citation statistics, and the share of studies using explainability). All bibliographic records derive from OpenAlex (metadata released under CC0); no proprietary or personal data are included. The search was capped at the 120 most relevant records per query (a large representative sample, not an exhaustive census).

Steps to reproduce

1. Search. Query the OpenAlex API (https://api.openalex.org/works) with twelve search strings combining AI/analytics, decision/prediction, and higher-education terms (full list in the article's Appendix A). Restrict to English-language journal and conference articles published 2015-2026 (filters: type:article, language:en, from_publication_date:2015-01-01, to_publication_date:2026-12-31). Retrieve the most relevant records per query (capped at 120/query). 2. Deduplicate. Merge results across the twelve queries and remove duplicates by OpenAlex work ID, yielding 1,190 unique records (corpus_raw.json). Keep the 1,105 records that have a retrievable abstract and a valid publication year for screening (corpus_screen.json). 3. Screen. Assess each record's title and abstract against the inclusion criteria (higher-education context; uses AI/ML/analytics; supports a decision via prediction, recommendation, early warning, or decision analytics; primary study) and exclusion criteria (not higher education; no AI/analytics; not decision support; off-topic; secondary review; non-research). Record for each item an include/exclude decision, a decision-support category, the principal AI method, a relevance rating (1-5), and a reason (screening_results.json, screening_final.json). 4. Build the funnel and core. Retain the 531 records meeting all inclusion criteria (included.json) and select a prioritized core of 95 studies by relevance and citation impact while preserving category balance (core.json). 5. Compute results. Derive the descriptive/bibliometric statistics (publication counts by year, decision-support categories, AI method distribution, author geography, venues, citation statistics, and the share of studies using explainability) directly from the included-study metadata (bibliometrics/stats.json). 6. References. Retrieve full bibliographic metadata for the 95 core studies from OpenAlex and add the verified background references (bib_core.json, bib_background.json). Note: counts may differ slightly on re-running because OpenAlex metadata are updated continuously. The search was conducted in June 2026.

Institutions

• Baku State University

  Baki, Baku

Categories

Licence